Splines were adjusted for demographics, comorbidities, blood pressure, heart rate, estimated glomerular filtration rate, and amino-terminal pro–B-type natriuretic peptide. The best fit for each model was achieved by using 3 knots for LV mass index (A); 4 knots for tissue Doppler e’ (D), E/e’ ratio (E), and left atrium (LA) volume index (F); and a linear model for left ventricular (LV) ejection fraction (B) and global longitudinal strain (C). The P values for trend (overall and linear) and P values for nonlinearity for each measure were as follows: A, overall P < .001 and P for nonlinearity = .005; B, linear P = .39; C, linear P = .59; D, overall P < .001 and P for nonlinearity = .001; E, overall P < .001 and P for nonlinearity = .002; and F, overall P < .001 and P for nonlinearity = .03.
Rates were based on the presence (+) or absence (–) of diastolic dysfunction (DD) and upper reference limit of hs-cTnT (high-sensitivity cardiac troponin T) concentration (TnT+).
aP < .05 vs DD- TnT- as reference.
bP < .05 vs DD- TnT+ as reference.
cP < .05 vs DD+ TnT- as reference.
eTable 1. Baseline Characteristics and Echocardiographic Measurements According to Categories of hs-cTnT at Visit 5 in the Total Population That Attended Visit 5 (i.e. Not Excluding Prevalent Cardiovascular Disease)
eTable 2. Echocardiographic Measurements According to Sex-Blinded Categories by Stratifying the Total Population in Quintiles of hs-cTnT at Visit 5, Irrespective of Sex
eTable 3. Baseline Characteristics Divided in Categories of hs-cTnT at Visit 5 of All Participants Alive at Visit 5 by Inverse-Probability-Weighted Estimation Analysis
eTable 4. Associations Between hs-cTnT (per 1 Log Unit Increase) at Visit 5 and Measures of Diastolic Dysfunction According to Validated Cut-offs Derived from the ARIC Study
eTable 5. Associations Between hs-cTnT (per 1 Log Unit Increase) at Visit 5 and Measures of Diastolic Dysfunction According to the Guidelines by American Society of Echocardiography
eTable 6. Association of Log-transformed hs-cTnT Concentrations at Visit 5 Adjusted for Age, Sex, Race, Hypertension, Diabetes, Obesity and eGFR and Incident Heart Failure (HF), HF With Reduced (HFrEF) or Midrange Ejection Fraction (EFmrEF) and HF With Preserved Ejection Fraction (HFpEF) Without and With Adjustment for Measures of Diastolic Function (Tissue Doppler e’, E/e’ and Left Atrial [LA] Volume index) and/or Left Ventricular [LV] Mass Index and LV Hypertrophy
eTable 7. Risk of Incident Heart Failure (HF), HF With Reduced (HFrEF) or Midrange Ejection Fraction (EFmrEF) and HF With Preserved Ejection Fraction (HFpEF) Stratified by the Presence (+) or Absence (-) of Diastolic Dysfunction (DD) and hs-cTnT > Upper Reference Limit (TnT+) After Adjusting for LV Mass Index
eTable 8. Risk Incident Heart Failure (HF), HF With Reduced (HFrEF) or Midrange Ejection Fraction (EFmrEF) and HF With Preserved Ejection Fraction (HFpEF) Stratified by the Presence (+) or Absence (-) of Diastolic Dysfunction (DD) Classified According to the American Society of Echocardiography (ASE) Guidelines and hs-cTnT > Upper Reference Limit (TnT+)
eTable 9. Incident Rates (per 100 Person-years) of Heart Failure (HF) Overall (114 Events), HF With MidRange or Reduced Ejection Fraction (HFmrEF/HFrEFm 50 Events), and HF With Preserved Ejection Fraction (HFpEF, 53 Events) Among Participant Categories Based on the Presence (+) or Absence (-) of Diastolic Dysfunction (DD) and hs-cTnT > Upper Reference Limit (TnT+)
eFigure. Left Ventricular Mass Index (g/m2), Mean Left Ventricular Wall Thickness (cm), Left Ventricular End Diastolic Volume Index (ml/m2), E/e’-ratio and LA Volume Index (ml/m2) Across Categories of hs-cTnT at Visit 5, Based on hs-cTnT Measurements at Visit 2 and Visit 4 (Marked as Non-detectable [ND] =<5ng/L and Detectable [D] = ≥5 ng/L)
Customize your JAMA Network experience by selecting one or more topics from the list below.
Myhre PL, Claggett B, Ballantyne CM, et al. Association Between Circulating Troponin Concentrations, Left Ventricular Systolic and Diastolic Functions, and Incident Heart Failure in Older Adults. JAMA Cardiol. 2019;4(10):997–1006. doi:10.1001/jamacardio.2019.3113
What is the association between high-sensitivity cardiac troponin T concentrations, left ventricular systolic and diastolic functions, and risk of heart failure?
In this analysis of a cohort of 4111 participants without cardiovascular disease, a greater concentration of high-sensitivity cardiac troponin T was associated with worse diastolic function but not with systolic function, independent of left ventricular mass. Left ventricular diastolic function accounted for a substantial proportion of the heart failure risk, and preserved ejection fraction was associated with greater high-sensitivity cardiac troponin T concentrations.
This study suggests that elevated troponin concentrations may serve as an early marker of subclinical alterations in cardiac structure and diastolic function that predispose a person to developing heart failure.
Cardiac troponin is associated with incident heart failure and greater left ventricular (LV) mass. Its association with LV systolic and diastolic functions is unclear.
To define the association of high-sensitivity cardiac troponin T (hs-cTnT) with LV systolic and diastolic functions in the general population, and to evaluate the extent to which that association accounts for the correlation between hs-cTnT concentration and incident heart failure overall, heart failure with preserved LV ejection fraction (LVEF; HFpEF), and heart failure with LVEF less than 50%.
Design, Setting, and Participants
This analysis of the Atherosclerosis Risk in Communities (ARIC) Study, an ongoing epidemiologic cohort study in US communities, included participants without cardiovascular disease (n = 4111). Available hs-cTnT measurements for participants who attended ARIC Study visits 2 (1990 to 1992), 4 (1996 to 1998), and 5 (2011 to 2013) were assessed cross-sectionally against echocardiographic measurements taken at visit 5 and against incident health failure after visit 5. Changes in hs-cTnT concentrations from visits 2 and 4 were also examined. Data analyses were performed from August 2017 to July 2018.
Main Outcomes and Measures
Cardiac structure and function by echocardiography at visit 5, and incident heart failure during a median 4½ years follow-up after visit 5.
Of the 6538 eligible participants, 4111 (62.9%) without cardiovascular disease were included. Among these participants, 2586 (62.9%) were female, and the mean (SD) age was 75 (5) years. Median (interquartile range) hs-cTnT concentration at visit 5 was 9 (7-14) ng/L and was detectable in 3946 participants (96.0%). After adjustment for demographic and clinical covariates, higher hs-cTnT levels were associated with greater LV mass index (adjusted mean [SE] for group 1: 33.8 [0.5] vs group 5: 40.1 [0.4]; P for trend < .001) and with worse diastolic function, including lower tissue Doppler imaging e’ (6.00 [0.07] vs 5.54 [0.06]; P for trend < .001), higher E/e’ ratio (11.4 [0.2] vs 12.9 [0.1]; P for trend < .001), and greater left atrial volume index (23.4 [0.4] vs 26.4 [0.3]; P for trend < .001), independent of LV mass index; hs-cTnT level was not associated with measures of LV systolic function. Accounting for diastolic function attenuated the association of hs-cTnT concentration with incident HFpEF by 41% and the association with combined heart failure with midrange and reduced ejection fraction combined (LVEF <50) by 17%. Elevated hs-cTnT concentration and diastolic dysfunction were additive risk factors for incident heart failure. For any value of late-life hs-cTnT levels, longer duration of detectable hs-cTnT from midlife to late life was associated with greater LV mass in late life but not with worse LV systolic or diastolic function.
Conclusions and Relevance
This study shows that higher hs-cTnT concentrations were associated with worse diastolic function, irrespective of LV mass, but not with systolic function; these findings suggest that high levels of hs-cTnT may serve as an early marker of subclinical alterations in diastolic function that may lead to a predisposition to heart failure.
Cardiac troponin measured by a high-sensitivity assay (hs-cTn) detects low-grade cardiomyocyte damage and is associated with incident heart failure and mortality in the general population1-5 as well as mortality and heart failure hospitalization in patients with heart failure.6-8 Increases in hs-cTn concentrations in 2 to 5 years are also associated with mortality among older adults in the community.4,9 Higher hs-cTn concentrations are associated with higher left ventricular (LV) mass in the general population2,10,11 and among patients with heart failure with preserved LV ejection fraction (LVEF; HFpEF).6,12 However, the association of hs-cTn levels with LV function, and diastolic function in particular, is largely uncharacterized. This situation is particularly true in older adults,13,14 among whom diastolic dysfunction prevalence, heart failure prevalence, and heart failure incidence are highest. Diastolic dysfunction shares many risk factors with LV hypertrophy, can occur in the absence of overt ventricular remodeling,15 and has a graded association with mortality and incident heart failure in the general population.16-19
We hypothesized that higher hs-cTnT (troponin T) concentrations are associated with worse diastolic function and that associated impairments in diastolic function partially account for the heightened risk of heart failure associated with higher hs-cTnT. We also hypothesized that increases in hs-cTnT levels from mid to late life would be associated with greater LV mass and worse diastolic function in late life. We tested these hypotheses in 4111 participants in the community-based Atherosclerosis Risk in Communities (ARIC) Study.
The ARIC Study is an ongoing epidemiologic cohort study that initially recruited 15 792 men and women aged 45 to 64 years in 4 US communities between 1987 and 1989 (visit 1).20 The concentration of hs-cTnT was measured at visits conducted in 1990 to 1992 (visit 2), 1996 to 1998 (visit 4), and 2011 to 2013 (visit 5). Echocardiography was performed at ARIC Study field centers during visit 5. The study protocol, including the present analysis, was approved by the institutional review boards at each field center, and all participants provided written informed consent.
For this present analysis, participants with prevalent heart failure (n = 965), coronary artery disease (n = 979), atrial fibrillation (n = 649), and/or stroke (n = 229) at visit 5 were excluded,21,22 along with those with missing echocardiography or hs-cTnT measurements (n = 463 [7.1% of the 6538 who attended visit 5]; among those who were alive at the start of visit 5, 3718 [36.2%] did not attend). We assessed the long-term change in hs-cTnT concentrations in participants who had available hs-cTnT measurements at visits 2, 4, and 5.
Comprehensive 2-dimensional Doppler, tissue Doppler imaging (TDI), and speckle-tracking echocardiography was performed at visit 5 at all sites using uniform imaging software and hardware by a predefined imaging protocol.23 Quantitative measures were performed according to recommendations by the American Society of Echocardiography (eMethods in the Supplement),24 and reproducibility metrics were excellent, as previously reported.23 Diastolic dysfunction was defined as abnormalities in 2 or more of TDI e′, E/e′ ratio, and left atrium (LA) volume index. Concentrations of hs-cTnT were measured with a high-sensitivity sandwich immunoassay from stored plasma samples (–80°C) drawn at visits 2, 4, and 5. The samples were analyzed at the University of Minnesota (Minneapolis) and Baylor College of Medicine (Houston, Texas) using high-sensitivity assay (Elecsys Troponin T; Roche Diagnostics) on an automated analyzer (Cobas e411; Roche Diagnostics).25 The limit of detection for hs-cTnT level was 5 ng/L, and the limit of blank was 3 ng/L. Details of assay performance are described in the eMethods in the Supplement. Unmeasurable concentrations of hs-cTnT were entered as 1.5 ng/L (half of the limit of blank). Amino-terminal pro–B-type natriuretic peptide (NT-proBNP) was measured at visit 5 on the automated analyzer.
Participants in the ARIC Study undergo active surveillance for incident cardiovascular events. Incident heart failure after visit 5 was based on ARIC Study committee adjudication of hospitalizations with International Classification of Diseases, Ninth Revision, and Tenth Revision, codes associated with heart failure, which includes abstraction of LVEF during that hospitalization if available, as previously described.26 Heart failure hospitalizations were classified as HFpEF if the LVEF was 50% or greater or as heart failure with midrange or reduced ejection fraction (HFmrEF or HFrEF) if the LVEF was less than 50% at the incident hospitalization; if LVEF was not assessed, then the most recent abstracted LVEF within 6 months of the index hospitalization was used. Participants were followed up for incident heart failure through December 31, 2016.
Participants were categorized into 5 groups according to hs-cTnT concentration at visit 5. Group 1 included participants with hs-cTnT below the limit of detection (<5 ng/L). Group 5 included participants with hs-cTnT levels above or equal to the sex-specific 99th percentile upper reference limit (14 ng/L for women and 22 ng/L for men) per the recent recommendations of the US Food and Drug Administration.27 Group 2 (5-7 ng/L for women and 5-9 ng/L for men), group 3 (8-9 ng/L for women and 10-14 ng/L for men), and group 4 (10-13 ng/L for women and 15-21 ng/L for men) were categorized by sex-specific tertiles of hs-cTnT concentrations between group 1 and group 5.
Demographic and echocardiographic variables were compared across hs-cTnT groups, and P for trend was calculated using linear or logistic regression with adjustment for age, sex, and race/ethnicity (model 1). For echocardiographic measures, further adjustment was made for body mass index, hypertension, diabetes, smoking, systolic blood pressure, heart rate, estimated glomerular filtration rate (eGFR) (based on the Chronic Kidney Disease Epidemiology Collaboration equation), and NT-proBNP measured at visit 5 (model 2). The covariates were selected by a priori knowledge. The continuous association between hs-cTnT levels and echocardiographic measurements was assessed with linear regression and with restricted cubic splines to investigate nonlinear associations, as detailed in the eMethods in the Supplement. Effect modification by sex and race/ethnicity was assessed with multiplicative interaction terms in linear regression models. Correlations were calculated by Spearman rank correlation.
The association between the presence of diastolic dysfunction, defined by ARIC Study–derived and American Society of Echocardiography–recommended cutoffs, and hs-cTnT concentration was assessed by multivariable logistic regression (model 2 covariates). The incremental value of hs-cTnT level in estimating diastolic dysfunction was assessed by the increase in the area under the receiver operating curve and the log likelihood from the addition of hs-cTnT concentration to a baseline logistic regression model that included all model 2 covariates (demographics, body mass index, hypertension, diabetes, smoking, systolic blood pressure, heart rate, eGFR, and NT-proBNP). To assess the potential implication of bias from selective visit 5 nonattendance among living cohort participants, we performed a sensitivity analysis using inverse probability of attrition weights (eMethods in the Supplement).
To quantify the extent to which associations with LV mass and diastolic function account for the association between hs-cTnT concentrations and incident heart failure (overall, HFpEF, and HFmrEF or HFrEF), we determined the proportional reduction in magnitude of the β coefficient for hs-cTnT in Cox proportional hazards regression models, adjusting for age, sex, and race/ethnicity, and after subsequent adjustment for (1) diastolic function (ie, TDI e′, E/e′ ratio, and LA volume), (2) LV mass, and (3) both diastolic function and LV mass. Associated 95% CIs were obtained from 2000 bootstrap replications using the percentile method. For time-to-event analyses with incident HFpEF as an outcome, participants with incident HFmrEF or HFrEF were censored at the time of heart failure and vice versa for incident HFmrEF or HFrEF.
The association of echocardiographic measures and change in hs-cTnT levels from visits 2 to 4 was assessed within each hs-cTnT concentration category (groups 1-5) at visit 5. Participants were divided in nondetectable (<5 ng/L) and detectable (>5 ng/L) hs-cTnT concentrations at visits 2 and 4. We calculated the trend between participants with stable low (nondetectable at both time points), increasing (nondetectable at visit 2 and detectable at visit 4), and stable high (detectable at both time points) as a measure of temporal exposure of hs-cTnT levels.
A 2-sided P < .05 was considered statistically significant. All statistical analyses were performed from August 2017 to July 2018 using Stata, version 13 (StataCorp LLC).
Of the 6538 eligible participants, 4111 (62.9%) without cardiovascular disease were included. Among these participants, 2586 (62.9%) were female, 3215 (78.2%) were white, and the mean (SD) age at visit 5 was 75 (5) years. A total of 6538 (63.8%) of 10 250 surviving participants attended visit 5. The median (interquartile range [IQR]) hs-cTnT concentration was 9 (7-14) ng/L and was detectable in 3946 participants (96.0%) but below the limit of blank (<3 ng/L) in 165 participants (4.0%). A higher group of hs-cTnT concentration was associated with older age, male sex, and black race/ethnicity as well as with a higher prevalence of cardiovascular comorbidities, higher body mass index and NT-proBNP, and lower eGFR after adjusting for age, race/ethnicity, and sex (Table 1).
Higher hs-cTnT group was associated with greater LV mass index (adjusted mean [SE] for group 1: 33.8 [0.5] vs group 5: 40.1 [0.4]; P for trend < .001), cavity size (adjusted mean [SE] left ventricular end-diastolic diameter for group 1: 4.28 [0.02] vs group 5: 4.42 [0.02]; P for trend < .001), wall thickness, and prevalence of hypertrophy (group 1: 8.4% vs group 5: 27.9%; P for trend < .001) (Table 1 and Figure 1), which remained statistically significant after full multivariable adjustment (model 2). Higher hs-cTnT groups were associated with worse LVEF (adjusted mean [SE] for group 1: 66.3 [0.3] vs group 5: 65.5 [0.5]; P for trend = .005) and global longitudinal strain (group 1: –18.3 [0.1] vs group 5: –17.8 [0.1]; P for trend < .001) in models adjusting for demographics (model 1) but were not statistically significant after further adjustment (Table 1 and Figure 1). Race/ethnicity and sex did not statistically significantly modify the association of hs-cTnT with LV mass index or with systolic measures.
Higher hs-cTnT group was associated with lower TDI e′ (adjusted mean [SE] for group 1: 6.00 [0.07] vs group 5: 5.54 [0.06]; P for trend < .001), higher E/e′ ratio (11.4 [0.2] vs 12.9 [0.1]; P for trend < .001), and higher LA volume index (23.4 [0.4] vs 26.4 [0.3]; P for trend < .001), which persisted in fully adjusted models, including NT-proBNP (Table 1). These associations were nonlinear, such that approximately linear associations were observed within the normal range of hs-cTnT level, whereas these associations were less robust at hs-cTnT concentrations above the normal range (Figure 1). No effect modification was noted by sex or race/ethnicity. The association of diastolic measures with hs-cTnT concentrations persisted after further adjustment for LV mass index (P < .001 for e', E/e' ratio, and LA volume index).
Similar associations were observed in the following sensitivity analyses: (1) including all participants with echocardiography and hs-cTnT measurement at visit 5 (ie, 6074 participants with prevalent cardiovascular disease not excluded; eTable 1 in the Supplement); (2) using groups defined by quintiles of hs-cTnT level in the overall analysis population, irrespective of sex (eTable 2 in the Supplement); and (3) incorporating inverse probability of attrition weights to account for possible bias from visit 5 nonattendance (eTable 3 in the Supplement).
Diastolic dysfunction was present in 761 participants (18.7%). In fully adjusted models (model 2), each logarithmic unit increase in hs-cTnT concentration was associated with 37% higher odds of diastolic dysfunction (odds ratio [OR], 1.37; 95% CI, 1.16-1.61; P < .001), and participants with hs-cTnT levels above the upper reference limit had 45% higher odds of diastolic dysfunction (OR, 1.45; 95% CI, 1.17-1.81; P = .001). Higher hs-cTnT concentration was also associated with abnormalities of each component diastolic measure (eTable 4 in the Supplement). Consistent results were observed when using the American Society of Echocardiography/European Society of Echocardiography criteria for classifying diastolic dysfunction (eTable 5 in the Supplement). Moreover, hs-cTnT measurement provided incremental value beyond established clinical risk factors in identifying participants with diastolic dysfunction as reflected in an increase in the area under the receiver operating curve by 0.90% (95% CI, 0.17%-1.62%) and the likelihood ratio test (266.4 vs 277.5; P < .001).
Incident heart failure occurred in 114 participants (2.8%) during the mean (SD) 4.5 (0.9) years of follow-up after visit 5. Higher hs-cTnT concentrations were associated with higher risk of incident heart failure overall (hazard ratio [HR], 2.58; 95% CI, 1.87-3.57) and incident HFpEF (HR, 2.23; 95% CI, 1.39-4.35) per log unit increase in hs-cTnT as well as with HFmrEF or HFrEF individually (HR, 2.65; 95% CI, 1.61-4.35 per log unit increase in hs-cTnT (Table 2). Adjusting for LV mass index and diastolic function together attenuated the association of hs-cTnT level with HFpEF by 55% and with HFmrEF or HFrEF by 47%. Adjusting for diastolic function attenuated the association of hs-cTnT concentration with incident HFpEF to a greater extent compared with the association of hs-cTnT level with HFmrEF or HFrEF (41% vs 17%). In contrast, adjusting for LV mass index attenuated the association of hs-cTnT concentration with incident HFmrEF or HFrEF to a greater extent compared with incident HFpEF (47% vs 23%; Table 2). These findings were consistent when also adjusting for hypertension, diabetes, obesity, and eGFR (eTable 6 in the Supplement).
Levels of hs-cTnT and diastolic dysfunction were independently associated with incident heart failure, and participants with both diastolic dysfunction and elevated hs-cTnT concentrations (above the upper reference limit or group 5) were at highest risk (incident rate, 2.76; 95% CI, 1.85-4.12) per 100 person-years (Figure 2). These findings were consistent for incident HFmrEF or HFrEF. However, elevated troponin only appeared associated with incident HFpEF among those with concomitant diastolic dysfunction. These results were consistent when also adjusting for LV mass index (eTable 7 in the Supplement) and when using guideline-recommended cutoffs for diastolic dysfunction (eTable 8 in the Supplement). Associations were attenuated when also adjusting for NT-proBNP, particularly for incident HFrEF, but remained consistent and statistically significant for incident heart failure overall and incident HFpEF (eTable 9 in the Supplement).
Among 3573 participants (86.9%) with hs-cTnT measurements at both visits 2 and 4, the median (IQR) hs-cTnT concentration at visit 2 was less than 3 (<3 to 4) ng/L and was below the limit of blank (<3 ng/L) in 2210 (61.9%) of 3573 participants. At visit 4, the median (IQR) hs-cTnT concentration was 4 (<3 to 6) ng/L and was below the limit of blank in 1502 (42.0%) of 3573 participants. The hs-cTnT level increased from visit 2 to visit 5 in 3396 (93.3%) of 3573 participants and from visit 4 to visit 5 in 3379 (90.1%) of 3573 participants.
Greater duration of detectable hs-cTnT level was associated with greater LV mass, volume, and wall thickness at visit 5 (eFigure in the Supplement). These associations were consistent across the hs-cTnT groups at visit 5 (P for interaction > .05 between hs-cTnT group at visit 5 and history of hs-cTnT at visits 2 and 4). For example, participants in the highest hs-cTnT groups at visit 5 with undetectable concentrations at visits 2 and 4 had LV mass comparable to those of participants with undetectable hs-cTnT concentration at visit 5. These results were consistent when replacing raw LV mass with LV mass indexed by body surface area but not when indexing to height (in meters) raised to the 2.7th power. In contrast, no consistent association was noted between duration of detectable hs-cTnT levels and measures of LV diastolic function at visit 5 (eFigure in the Supplement).
Among 4111 older adults without cardiovascular disease, higher circulating concentrations of hs-cTnT were independently associated with alterations in LV structure (greater mass, volume, wall thickness, and concentricity) and worse diastolic function (TDI e′, E/e′ ratio, and LA volume) but not with measures of systolic function. The associations between hs-cTnT levels and diastolic function were independent of LV mass, and higher hs-cTnT concentrations were associated with the severity of diastolic dysfunction. Adjusting for diastolic function and LV mass together resulted in an attenuation of the association of hs-cTnT level with incident heart failure by approximately 46%. Diastolic measures attenuated the association of hs-cTnT concentration with incident HFpEF to a greater extent than incident HFmrEF or HFrEF, whereas LV mass attenuated the association of hs-cTnT concentration with HFmrEF or HFrEF to a greater extent. Participants with both diastolic dysfunction and elevated hs-cTnT concentration were at a particularly high risk of heart failure. For any value of late-life hs-cTnT measurement, longer duration of detectable hs-cTnT levels from midlife to late life was associated with greater LV mass in late life but not with worse LV function.
These results from a community-based older adult cohort extend the findings of previous studies that hs-cTnT level is associated with measures of LV structure and the risk of incident heart failure overall (Table 3). The key novel finding of the present analysis is the robust association between hs-cTnT levels and measures of diastolic function in these older adults, including TDI e′, E/e′ ratio, and LA size, particularly within the normal range of hs-cTnT concentration. Although the unadjusted association of hs-cTnI (troponin I) level with E/e′ ratio has been previously demonstrated in a middle-aged cohort (mean [SD] age, 55  years) with generally normal diastolic indices,13 the present study included participants with a higher prevalence of abnormal diastolic indices; integrated 3 different measures to assess diastolic function; and adjusted for cardiovascular risk factors, NT-proBNP, and LV mass.
Our findings are consistent with results from studies in isolated cardiomyocytes and transgenic mice, suggesting an association between cardiac troponin and impaired diastolic, but not systolic, function.28,29 Recent research that used invasive hemodynamic assessments of patients with HFpEF at rest and during exercise suggests that elevated LV filling pressures and subendocardial ischemia from impaired myocardial oxygen supply-demand balance are 2 factors in troponin elevation in HFpEF.30 The lack of serial echocardiography in the present study precludes the identification of the temporal association between troponin elevation and diastolic dysfunction, but our findings suggest that these observations extend to older adults without, but are at risk of, HFpEF. Alterations in diastolic function accounted for a sizeable proportion of the association between hs-cTnT level and incident HFpEF in this analysis. The mechanisms linking diastolic dysfunction to elevated troponin and driving oxygen supply-demand mismatch are likely multifold. Although higher hs-cTnT concentrations are independently associated with inducible ischemia among patients with stable coronary artery disease,31 data from patients without overt coronary artery disease suggest an important role for coronary microvascular ischemia.32 Coronary microvascular dysfunction is common in HFpEF, even in the absence of unrevascularized epicardial coronary artery disease.33 Further studies are necessary to define the extent to which microvascular disease mediates the association between diastolic dysfunction and hs-cTnT level in older adults.
Previous studies have demonstrated the associations between cardiac troponin and systolic indices (LVEF and mitral annular plane systolic excursion) in participants with heart failure.34,35 In this study’s older adult population without cardiovascular disease and with largely preserved LVEF, higher hs-cTnT concentration was associated with worse LVEF and global longitudinal strain in unadjusted analysis, which is similar to findings from the Cardiovascular Health Study.4 However, these associations did not persist after accounting for cardiovascular risk factors. These findings are concordant with recent results from the Multi-Ethnic Study of Atherosclerosis, which demonstrated hs-cTnT to be a marker of increasing LV mass and LV dilatation but not declining systolic function (measured by cardiac magnetic resonance with a 10-year interval) after covariate adjustment in the general population.10 Participants with prevalent heart failure, and thus most participants with reduced LVEF, were also excluded in this study. Although higher hs-cTn level has been associated with worse systolic function in patients with a more advanced condition, including chronic kidney disease, findings from the present study suggest greater sensitivity of hs-cTnT concentration for alterations in LV mass and diastolic function compared with systolic performance among older adults without cardiovascular disease among whom LVEF is largely preserved.
The prognostic value of hs-cTn for mortality and incident heart failure in the general population has been extensively demonstrated,1,2,4 including in the ARIC Study.3 Results of this study suggest that diastolic function accounts for a substantial proportion of the risk of incident heart failure, particularly the risk of incident HFpEF, identified by hs-cTnT measurement in this older population. Hence, elevated hs-cTnT levels may serve as an early marker of subclinical alterations in cardiac structure and diastolic function that predispose individuals to the development of heart failure. Similar to previous findings of additive risk for heart failure between greater LV mass and hs-cTnT concentration,36 we found that participants with both diastolic dysfunction and elevated hs-cTnT level were at statistically significantly higher risk compared with those with only 1 of the 2 present. This pattern was also observed for incident HFpEF and HFmrEF or HFrEF, although our power was limited in these subgroups. Echocardiographic measures of LV structure and function may, therefore, be especially relevant to prognosis in persons with elevated troponin, particularly for incident HFpEF,37 which accounts for most of the heart failure events in older people.38
During the 2 decades preceding visit 5, hs-cTnT concentrations increased in most participants (>90%). Because most participants had unmeasurable hs-cTnT levels in midlife, the increase was highly correlated with the concentration in late life. Despite this, greater duration of hs-cTnT exposure was associated with measures of LV structure beyond the late-life hs-cTnT concentration, including greater LV mass, volume, and wall thickness. In contrast, greater duration of hs-cTnT exposure was not statistically significantly associated with metrics of LV diastolic function beyond the visit 5 value, suggesting that the duration of detectable hs-cTnT level is particularly relevant to LV structure but not to diastolic function.
This study has several limitations. We used noninvasive measures of diastolic function because cardiac catheterization is not feasible in a large community-based cohort and diastolic function is most commonly assessed using echocardiography. Although degradation of frozen hs-cTnT samples may cause preanalytical variability, previous data from the ARIC Study show no statistically significant bias when comparing measurements at the different time points.39 Among ARIC study participants who were alive at the start of visit 5, 3718 (36.2%) did not attend this visit, and some were excluded from the analysis owing to missing echocardiography or hs-cTnT measurements (n = 463 (7.1%) of 6538 who attended visit 5). Sensitivity analysis using inverse probability of attrition weights demonstrated similar associations as the primary analysis (eTable 3 in the Supplement).
This study found that, among older adults in a community-based population without overt heart disease, hs-cTnT concentrations were significantly associated with LV diastolic function but not with systolic function, independent of LV mass. Greater duration of detectable hs-cTnT level from mid to late life was associated with worse LV structure but not with diastolic function in late life even after accounting for late-life concentrations. Adjusting for measures of diastolic function and LV mass together attenuated the association of hs-cTnT concentration with incident heart failure by nearly half, with diastolic dysfunction attenuating a relatively larger proportion of the association of hs-cTnT with HFpEF. The presence of both elevated hs-cTnT level and diastolic dysfunction was associated with a particularly high risk of incident heart failure. Elevated troponin may, therefore, serve as an early marker of subclinical alterations in diastolic function that predispose one to heart failure development.
Accepted for Publication: July 15, 2019.
Corresponding Author: Amil M. Shah, MD, MPH, Division of Cardiovascular Medicine, Brigham and Women’s Hospital, 75 Francis St, Boston, MA 02445 (email@example.com).
Published Online: September 4, 2019. doi:10.1001/jamacardio.2019.3113
Author Contributions: Dr Shah had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Myhre, Solomon, Skali, Shah.
Acquisition, analysis, or interpretation of data: Myhre, Claggett, Ballantyne, Selvin, Rosjo, Omland, Skali, Shah.
Drafting of the manuscript: Myhre, Shah.
Critical revision of the manuscript for important intellectual content: Myhre, Claggett, Ballantyne, Selvin, Rosjo, Omland, Solomon, Skali.
Statistical analysis: Myhre, Claggett, Skali, Shah.
Obtained funding: Shah.
Administrative, technical, or material support: Myhre, Shah.
Supervision: Rosjo, Omland, Solomon, Skali, Shah.
Conflict of Interest Disclosures: Dr Myhre reported receiving speaker honoraria from Novartis outside the submitted work. Dr Ballantyne reported receiving grants and personal fees from Roche and Abbott during the conduct of the study as well as being a holder of a provisional patent (No. 61721475; titled “Biomarkers to Improve Prediction of Heart Failure Risk”) filed by Baylor College of Medicine and Roche. Dr Røsjø reported receiving speaker honoraria from Novartis as well as personal fees from CardiNor AS and Thermo Fisher Brahms outside the submitted work; he reported holding a patent to Secretoneurin as a cardiac biomarker. Dr Omland reported receiving consultancy and/or speaker honoraria from Abbott Diagnostics, Roche Diagnostics, and Novartis; research support from AstraZeneca, Abbott Diagnostics, Novartis, Roche Diagnostics, Thermo Fisher, Singulex, and Biomedica via Akershus University Hospital; personal fees from Siemens; and grants from Somalogic outside the submitted work. Dr Solomon reported being a consultant for Roche Diagnostics; receiving grants from Alnylam, Amgen, AstraZeneca, Bellerophon, Bayer, BMS, Celladon, Cytokinetics, Eidos, Gilead, GSK, Ionis, Lone Star Heart, Mesoblast, MyoKardia, NIH/NHLBI, Novartis, Sanofi Pasteur, Theracos; and receiving personal fees from Akros, Alnylam, Amgen, AstraZeneca, Bayer, BMS, Cardior, Corvia, Cytokinetics, Gilead, GSK, Ironwood, Merck, Myokardia, Novartis, Roche, Takeda, Theracos, Quantum Genetics, Cardurion, AoBiome, Janssen, Cardiac Dimensions, and Tenaya outside the submitted work. Dr Shah reported receiving research support from Novartis; consulting fees from Philips Ultrasound and Bellerophon; and grants from National Institutes of Health/National Heart, Lung, and Blood Institute during the conduct of the study. No other disclosures were reported.
Funding/Support: The Atherosclerosis Risk in Communities Study is funded in whole or in part by contracts HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, and HHSN268201700004I from the National Heart, Lung, and Blood Institute (NHLBI), the National Institutes of Health (NIH), and the US Department of Health and Human Services. This study was supported by grants K08HL116792, R01HL135008, and R01HL143224 from the NHLBI (Dr Shah); the Watkins Discovery Award from the Brigham and Women’s Heart and Vascular Center (Dr Shah); grants K24DK106414 and R01DK089174 from the NIH/National Institute of Diabetes and Digestive and Kidney Diseases (Dr Selvin); grant R01HL134320 from the NIH/NHLBI (Drs Ballantyne and Selvin); and research grant from the South-Eastern Norway Regional Health Authority, the Norwegian Medical Association, and the Unger Vetlesen Medical Fund (Dr Myhre). The reagents for the hs-cTnT assays were donated by Roche Diagnostics.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank the staff and participants of the ARIC Study for their important contributions.